findFundamental() confusion

I got confused when I use findFundamental() function to obtain the fundamental matrix. I am not sure: whether it use all inliars to calculate the final fundamental matrix again or not.

In the function desctiption in calib.hpp said yes;
but when I go deep into fundam.cpp and ptstreg.cpp, I found that it just return the best model, and doesn’t use all inliars to calculate again.

Should I believe the code or the description?

Could you point to the corresponding code line on GitHub - opencv/opencv: Open Source Computer Vision Library?

Anyway, the code is what is executed ultimately.